#human-evaluations

#human-evaluations

The /insights command analyzes the history of your commands in the last 30 days and generates a detailed HTML report with key findings and insights.

For decades in SAAS, products reduced ambiguity. Users supplied constrained inputs, and the system handled the output. It's never been Minority Report cinematic, but it was predictable. By providing predictable environments for manipulating data, users learned by moving things, adjusting variables - and the outcome emerged through interaction.

Performance is a critical factor in user engagement, where even minor delays in loading can deter users. A clean and simple user interface also contributes significantly to user retention.

AI is disrupting more than the software industry, and is doing so at a breakneck speed. Not long ago, designers were deep in Figma variables and pixel-perfect mockups. Now, tools like v0, Lovable, and Cursor are enabling instant, vibe-based prototyping that makes old methods feel almost quaint. What's coming into sharper focus isn't fidelity, it's foresight. Part of the work of Product Design today is conceptual: sensing trends, building future-proof systems, and thinking years ahead.

My role was straightforward: write queries (prompts and tasks) that would train AI agents to engage meaningfully with users. But as a UXer, one question immediately stood out - who are these users? Without a clear understanding of who the agent is interacting with, it's nearly impossible to create realistic queries that reflect how people engage with an agent. That's when I discovered a glitch in the task flow. There were no defined user archetypes guiding the query creation process. Team members were essentially reverse-engineering the work: you think of a task, write a query to help the agent execute it, and cross your fingers that it aligns with the needs of a hypothetical "ideal" user - one who might not even exist.

Your junior designer spins up a prototype in Lovable before lunch. Your PM shows you a "working" MVP built entirely with Cursor within a day. And your CEO forwards you a LinkedIn post about how AI will replace 80% of UI work by 2026. And it seems like anyone can now make an app to solve a specific problem. Has the graphical interface really died, as Jakob Nielsen provocatively suggests?

The question dropped into the Slack channel before the user research summary. Before the problem was clearly defined. Before anyone asked if users actually needed this feature. Your product manager already generated three interface options in ChatGPT. Now they're asking which one to build. Not whether to build. Not why to build. Which. And when you slow the conversation down to ask those questions, you're about to discover that strategic thinking now reads as bottleneck behavior.

Progressive disclosure is a well-known principle in UX design. This principle is about showing users only what they need right now, and revealing more options or information gradually as they interact or gain context. The goal is to reduce cognitive load, keep interfaces clean and approachable, and still support advanced use cases when needed. The principle of progressive disclosure can be applied not only to the user interfaces we design, but also AI tools we use.